techiescamp/mlops-for-devops

What it solves

Most MLOps tutorials assume you already know data science. This project flips that: it teaches Machine Learning Operations to DevOps, Platform, and SRE engineers who have no ML background. Instead of asking you to become a data scientist, it explains every ML concept using DevOps analogies you already understand, so you can build, automate, deploy, and troubleshoot ML systems in production just like you do with conventional applications.

How it works

It’s a project-based roadmap split into two phases. Phase 1 starts on your local machine: you build an end-to-end employee attrition prediction model for a fictional 500,000-person company, covering dataset pipelines, data preparation, model training with scikit-learn, and serving the model as a live API using KServe. Phase 2 moves to enterprise orchestration: you replace manual steps with automated, scalable workflows on Kubernetes using tools like DVC for data versioning, Airflow for pipeline orchestration, Feast for feature stores, Kubeflow for training pipelines, MLflow for experiment tracking, KServe for model serving, and Evidently AI for monitoring. Everything runs on Kubernetes and Docker. There are linked step-by-step guides for each tool, and the repo contains the code and configuration manifests for both phases. A capstone project is planned that combines everything into a production-style MLOps platform on Kubernetes.

Who it’s for

DevOps, Platform, and SRE engineers who want to operate ML workloads without becoming data scientists. It’s also useful for anyone who understands infrastructure and wants a hands-on, project-driven introduction to MLOps using the same Kubernetes-based tools used in real-world platforms.

Highlights

• No ML background required — every concept is explained through DevOps analogies
• Uses one consistent use case (employee attrition prediction) end-to-end, so you focus on infrastructure, not ML theory
• Covers the full MLOps lifecycle: data pipelines, training, serving, experiment tracking, versioning, feature stores, and monitoring
• Entirely Kubernetes and Docker based, using industry-standard open-source tools like Airflow, Kubeflow, MLflow, KServe, DVC, Feast, and Evidently AI
• Includes code and configuration for both local development and enterprise orchestration phases
• Curated list of real-world MLOps reading materials and certification links for further learning